Using Bayesian Spatial Models to Facilitate Water Quality Monitoring

• Activity Number: 81
• Type: Contributed
• Date/Time: Monday, August 4, 2003 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics & the Environment
• Author(s): Steven S. Carroll*+ and Anthony R. Olsen
• Companies: Oregon State University and U.S. Environmental Protection Agency
• Keywords: spatial interpolation ; Bayesian models

Abstract

The Clean Water Act of 1972 requires states to monitor the quality of their surface water. The number of sites sampled on streams and rivers varies widely by state. A few states are now using probability survey designs to select sites, while most continue to rely on other procedures for site selection. When reporting on their waters, states must provide state-wide estimates on the number of stream miles that are impaired and list all impaired streams and rivers. The latter requirement, as well as a desire to provide water quality information for subregions of the state, is difficult to meet with typical sample sizes available. We propose the application of Bayesian spatial modeling to interpolate water chemistry measures at sites with no observed data. We show how Bayesian hierarchical models can be used for spatial interpolation. Then we use the interpolated values to estimate a spatial cumulative distribution function for assessing the water quality in a subregion. We illustrate the approach using data collected on Strahler first, second, and third order streams located in the mid-Atlantic region. The data were collected by the U.S.Environmental Protection Agency.